To understand the effects of financial incentives, health, and other factors on retirement decisions of individuals who live in couples, it is important to take account of joint features of their retirement-related decisions. Different paradigms about how couples make decisions lead to different predictions but also different welfare implications, implying the need for a better understanding of how households actually make decisions. This application aims at collecting new and innovative stated preference (SP) data, based on responses to choice questions featuring hypothetical retirement patterns. These data will help deepen our understanding of the black box of intra-household allocation decisions over what we are capable to infer from existing survey data. [unreadable] [unreadable] We propose to estimate various models of household decision-making, in which the household is assumed to make Pareto optimal decisions, using stated-preference data collected from respondents in the American Life Panel, an Internet panel that is largely representative of the age 40+ US population. This additional information is needed since it is well-known that collective models featuring complementarity in leisure and/or public goods will in general not be identified from observed actual retirement choices alone. Respondents will be asked a number of times to choose among different alternatives featuring different hours of work and income replacement ratios for both spouses. They will be asked explicitly to consider their own preferences, their spouse's preferences, and the most likely outcome the household would choose. This application shows how this innovative type of information identifies uniquely a set of individual specific preferences along with a bargaining weight function which nests as special cases the unitary paradigm in which the household acts as a single decision maker. This application also extends on the literature on intertemporal decision making of households by allowing households to be forward looking, incorporating the role of savings as an extension to the baseline model. [unreadable] [unreadable] Estimated preferences (and their distribution in the population) will then be applied to respondents in the Health and Retirement Study, a longitudinal survey representative of the U.S. elderly population, to evaluate the capacity of the SP based models to fit observed patterns of retirement. The MRS offers rich information on retirement prospects of respondents along with matched earnings histories from Social Security records and pension data provided by employers that allow to carefully assess opportunities that workers face when making retirement decisions. We will also study how policy changes affect retirement decisions and welfare [unreadable] [unreadable] [unreadable]